We’d like to offer our users the ability to run Jupyter in OOD from one of their own conda environments. I saw a few ideas here that (I think) could work for us:
But we were interested if nb_conda_kernels could let our users switch envs in Jupyter OOD. Our thought was that the user would startup Jupyter in OOD from our sitewide module (as usual), and then they could switch to a Python kernel from their own conda env. Has anyone tried setting this up?
If not, I’ll give it a shot a shot and report back here.
Research Scientist II
Partnership for an Advanced Computing Environment (PACE)
Georgia Institute of Technology
Welcome to the board. Thanks for posting! One of my colleagues referenced the following link. Please let us know if it helps or not.
Thanks Gerald! For this to work, do we need to do any extra configuration for our Jupyter app in OOD?
I would think that, in our
script.sh.erb, we would have the logic to run the user’s jupyter’s kernel, rather than our sitewide jupyter kernel. If so, do you have an additional input field for the user’s desired environment?
From an OOD perspective it’s just a module so there’s nothing on that side.
The module itself that get’s loaded may have to do some fancy footwork. @zyou How is our Jupyter module setup such that it can load a users’ kernels?
This worked really well for us:
If the user creates an ipython kernel from their virtual env (using
ipykernel install), then Jupyter will pick it up. Then their kernel will be available from the “New” dropdown menu along with the other pre-installed kernels. We didn’t have to reconfigure anything in our OOD setup.
Hello! We are have a custom environment option for users at the University of Central Oklahoma.
You can find our our Jupyter OnDemand application here: GitHub - UCO-HPC/buddy-jupyter: An app designed for UCO's OnDemand that launches a Jupyter server session
This method has worked well for us so far. Let me know if there are any questions about our setup.